fix: prevent false negatives in model catalog research

[Erin McNulty (erin2722)]

Summary

• require direct checks of the canonical OpenAI, Anthropic, and Gemini documentation before rejecting a requested model
• clarify that search results and omissions from secondary official pages are not evidence of unavailability
• require the agent report to record every official URL checked, exact slug evidence, pricing provenance, inference-surface fit, and any unresolved blocker

The existing catalog edit scope, official-source requirements, pricing rules, and fail-closed behavior remain unchanged.

Validation

• pnpm exec prettier --check .github/workflows/add-model.yaml
• parsed .github/workflows/add-model.yaml with Ruby YAML
• git diff --check origin/main...HEAD

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated
ai-proxy	Ready	Preview	Sep 23, 2026 2:44pm UTC

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-09-23T14:46:46.601185Z	473307e	PR opened

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "Codex (@codex) review" or "Codex (@codex) security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.